> [!NOTE] > **Note:** The Agentic UI toolkit is in the experimental (pre-GA) stage. Pre-GA products and features might have limited support, and changes to them might not be compatible with other pre-GA versions. Pre-GA offerings are covered by the [Google Maps Platform Service Specific Terms](https://cloud.google.com/maps-platform/terms/maps-service-terms). For more information, see the [launch stage descriptions](https://developers.google.com/maps/launch-stages.md.txt). There's no charge for using the toolkit while it's in the experimental stage, but it might call other Google Maps Platform tools or services that incur a charge. ## Overview The Maps Platform **Agentic UI toolkit** is a developer toolkit designed specifically for the era of conversational AI. It provides a set of LLM system instructions and a library of UI components that help large language models (LLMs) and AI agents build real-time, interactive visual responses for geospatial data. Instead of hard-coding visual elements, you can use the toolkit to translate "intent-based instructions" from an LLM into dynamic, context-aware maps and place displays, keeping users engaged within the chat experience. Common use cases include the following: - **AI assistants**: Augmenting text responses with visual context for local queries, such as hotels and restaurants. - **Travel planners**: Showing maps, routes, and place details inline. - **Low-code development**: Reducing the friction of manually orchestrating multiple Maps APIs across web and mobile touchpoints. ## Key concepts ### 1. Agent-ready UI Create scalable, high-quality visuals---like place cards, immersive map views, and route previews---specifically designed for real-time LLM-powered experiences. ### 2. Context-rich experiences Help users discover and decide without leaving the chat. Provide rich inline maps context, such as photo galleries, ratings, or route options, to keep users engaged and give them the confidence to take action. ### 3. Intelligent components The Agentic UI toolkit handles complex geospatial logic automatically, instantly adapting context and layout to user intent while styling to ensure every interaction feels connected to your brand. ### 4. LLM-agnostic The Agentic UI toolkit is model-agnostic to your AI stack. Whether you use Gemini, models from OpenAI or Anthropic, and regardless of your grounded data source, the toolkit provides a dedicated presentation layer to visualize location data. ### 5. Server-side template architecture Server-side template architecture speeds up component rendering by performing template expansion and parameter resolution on the server side. This improves latency for real-time conversational experiences. ## Get started To start using the Agentic UI toolkit, you need to initialize it with a Google Gemini API key and a Google Maps Platform API key. For more information, see [Get started](https://developers.google.com/maps/ai/agentic-ui-toolkit/getting_started.md.txt). ## UI outcomes The Agentic UI toolkit provides a comprehensive suite of specialized UI abilities out of the box. These components are available across the web, Android, and iOS. - **Place Detail, compact**: Specific point of interest (POI) information in a compact card - **Inline Maps**: Point or area locator - **Inline Map + Route**: Navigation or journey previews - **Inline Map Detail**: Location details or context using rich imagery ## System instructions The Agentic UI toolkit provides "system instructions" to guide the host LLM on when and how to construct the UI. ### Guidelines The host model must understand the following concepts: - How to match components to user intent, such as using Inline Map for orientation queries. - How to map parameters from natural language to schema, such as setting the map mode to satellite if the user asks for "scenic." ### Tool use schema The Agentic UI toolkit exposes a catalog of definitions compatible with standard LLM tool use schema, such as function calling. ## Policies and Terms of Service This library uses Google Maps Platform services. Use of Google Maps Platform services through this library is subject to the Google Maps Platform [Terms of Service](https://cloud.google.com/maps-platform/terms). Some APIs and backend services, such as [Grounding Lite](https://developers.google.com/maps/ai/grounding-lite/attribution) and [Grounding with Google Maps](https://docs.cloud.google.com/vertex-ai/generative-ai/docs/grounding/grounding-with-google-maps#google-maps-attribution-guidelines), have additional Terms of Service requirements that must also be met. This library isn't a Google Maps Platform Core Service. Therefore, the Google Maps Platform Terms of Service---such as Technical Support Services, Service Level Agreements, and the Deprecation Policy---don't apply to the code in this library. ### Support This library is offered through an open-source license. It isn't governed by the Google Maps Platform Technical Support Services Guidelines, the SLA, or the [Deprecation Policy](https://cloud.google.com/maps-platform/terms). However, any Google Maps Platform services used by the library remain subject to the Google Maps Platform Terms of Service. ### Attribution requirements for Google Maps sources Each tool response from the Agentic UI toolkit includes sources. When presenting results that use tools provided by the toolkit, you must include the associated Google Maps sources in a way that meets the following requirements: - The Google Maps sources must immediately follow the generated content that the sources support. - The Google Maps sources must be viewable within one user interaction. ## Share feedback To share feedback about the Maps Agentic UI toolkit, use the following forms: - [Report a bug](https://issuetracker.google.com/issues/new?component=2130972&template=2323416) - [Submit a feature request](https://issuetracker.google.com/issues/new?component=2130972&template=2323469)